> derided researchers in machine learning who use purely statistical methods to produce behavior that mimics something in the world, but who don't try to unders

> derided researchers in machine learning who use purely statistical methods to produce behavior that mimics something in the world, but who don't try to understand the meaning of that behavior. It's crazy how wrong Chomsky was about machine learning. Maybe the real truth is that humans are stochastic parrots who have an underlying probability distribution - and because gradient descent is so good at reproducing probability distributions - LLMs are incredibly good at reproducing language.